Sovereign AI in the Developing Global South: Architecting Strategic Autonomy in the Intelligence Economy

Artificial intelligence is turning the older digital divide of the developing world into an intelligence-economy participation divide: an asymmetry in the capacity of states to shape, own, govern and draw value from cognitive automation. Read against the literatures on digital colonialism, data extractivism and the "Data Nullius" framework, this article treats Sovereign AI as a distinct policy object for the Global South rather than a subset of digital public infrastructure or industrial policy. It sets out four interlocking registers of sovereignty (compute, supply chain, model and data) and argues for a modular, developmental architecture governed through "fluid institutionalism", built on the principle of foundations before the frontier: connectivity, device affordability and FAIR-aligned digital public infrastructure first, then retrieval-grounded small language models. The article then examines whole-of-government interoperability and tiered human-in-the-loop governance; economic transformation through public procurement; security resilience through inbound and outbound data embassies; local-language models such as Bhashini and InkubaLM; a "choose partners, not sides" posture in a bipolar technology rivalry; and South-South cooperation for a stronger voice in global AI governance. It closes with the binding constraints of energy, water, fiscal capacity and coordination, an assessment of risks including dependency lock-in, elite capture and linguistic exclusion, and a time-phased set of recommendations by institutional actor.